Pith. sign in

REVIEW 3 major objections 5 minor 30 references

MTVDiff: Multimodal Conditional Latent Diffusion for Enhanced Thermal-to-Visible Face Translation

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Pith's one-line read Multimodal latent diffusion with depth and text guidance outperforms prior methods in thermal-to-visible face translation.

desk verdict Plausible engineering with a solid ablation story, but the SOTA claim rests on an unstated assumption about what Depth Anything and LLaVA are given—if they see the ground-truth visible face, the comparisons are unfair. read the letter →

arxiv 2607.19886 v1 pith:XLVH72UP submitted 2026-07-22 cs.CV

classification cs.CV
keywords thermal-to-visiblefacetranslationlatentdiffusionmodelmultimodalfusioncross-spectralrecognitiondepth-guidedgenerationtext-guidedidentitypreservation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Thermal-to-visible face translation converts infrared face images into visible-light photos, a hard problem because thermal images lack color, texture, and geometric detail. The paper proposes MTVDiff, a latent diffusion model that conditions generation on three inputs: the thermal image, a depth map, and a text description of the person. On the MCXFace and SpeakingFaces benchmarks, MTVDiff claims to beat prior GAN- and diffusion-based methods, with FID reduced by up to 48.3% and face-verification Rank-1 accuracy improved by up to 8.9 percentage points. If these results hold, the method is a step toward practical nighttime face recognition using thermal sensors. The paper attributes the gains to three modules: dual-branch cross-attention fusion for thermal-depth features, gated text-to-visual alignment for semantic guidance, and spatial feature transformations that inject multimodal information into a frozen diffusion backbone.

What carries the argument

The framework builds on a frozen Stable Diffusion latent diffusion model. A Dual-Branch Cross-Attention Fusion (DBCAF) module runs two ResNet-18 encoders over thermal and depth inputs, fusing multi-scale features through cross-attention and learned channel weights. A Gated Text-to-Visual Feature Alignment layer inserts gated cross-attention and feed-forward blocks that blend CLIP text embeddings with the fused visual features, using learnable tanh-gated scalars to regulate influence. Spatial Feature Transformations (SFT) compute per-layer scale and bias from the fusion features and modulate the residual blocks of the frozen UNet. Together these let the model inject structural and semantic co

What would settle it

Run the same Depth Anything and LLaVA pipeline on the thermal input instead of the visible face, retrain MTVDiff, and compare FID and Rank-1 against the reported numbers. If the gap collapses, the depth/text conditions were leaking target content; if the gap persists, the multimodal design is genuinely responsible.

Watch

Extended reading notes

Core claim

The central claim is that adding two auxiliary conditions — depth maps and text captions — to a latent diffusion model substantially improves thermal-to-visible face translation. On MCXFace, MTVDiff achieves FID 75.33, SSIM 0.8355, and Rank-1 accuracy 87.26%; on SpeakingFaces, FID 14.37 and Rank-1 93.76%, beating the best prior method DiffV2IR by 48.3% in FID and the best overall baseline by 7.17 percentage points in Rank-1. The paper also shows via ablations that depth contributes the largest single-modality gain, that text only helps when combined with depth, and that a thermal-only variant still beats most baselines.

Load-bearing premise

The paper assumes that depth maps and text descriptions given to the model at test time are derived from the ground-truth visible face rather than from the thermal input; this is never stated explicitly, and if it were true it would mean the model is being conditioned on information from the very target it is supposed to reconstruct.

Editorial extensions

If this is right

  • Nighttime face recognition could work from thermal sensors alone, since generated visible faces are more recognizable by standard face matchers.
  • The framework is robust to missing modalities: a thermal-only variant still beats most baselines, so deployment can fall back when depth or text is unavailable.
  • Explicit depth and text conditioning offers controllability — attributes like age or gender can be specified at generation time.
  • Because the Stable Diffusion backbone is frozen, the trainable modules are relatively small, which may transfer to other cross-spectral image synthesis tasks.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The reported gains may partly come from the depth and text signals being derived from the ground-truth visible face, since the paper does not state what images Depth Anything and LLaVA are run on; if so, a fair comparison to thermal-only baselines would need those baselines to receive equivalent target-derived information.
  • A testable extension is to condition on depth and text estimated from the thermal input itself, which would make the pipeline self-contained and reveal how much of the gain is genuinely due to cross-modal reasoning rather than leakage of target content.
  • The prompt-sensitivity analysis suggests an automatic quality estimator for text prompts could further stabilize generation when descriptions are noisy, a direction the authors flag as future work.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes MTVDiff, a latent diffusion framework for thermal-to-visible face translation that conditions on thermal imagery, depth maps, and text prompts. Three components are introduced: a Dual-Branch Cross-Attention Fusion (DBCAF) module for thermal–depth fusion, a Gated Text-to-Visual Feature Alignment mechanism, and Spatial Feature Transformations (SFT) for injecting multimodal features into a frozen Stable Diffusion backbone. Experiments on MCXFace and SpeakingFaces report state-of-the-art results, e.g., FID 75.33 and Rank-1 87.26% on MCXFace, and FID 14.37 and Rank-1 93.76% on SpeakingFaces, with claimed FID reductions up to 48.3% and Rank-1 gains up to 8.9 percentage points over existing baselines. The paper includes ablations isolating depth, text, and cross-attention, a text-prompt sensitivity analysis, and a missing-modality robustness study.

Significance. If the experimental protocol is sound, MTVDiff would be a meaningful advance in cross-spectral face translation: the modular design is coherent, the ablations are systematic, and the gating-parameter analysis in §4.6 provides a useful diagnostic. The reported gains are large and falsifiable on public benchmarks, and the authors include a thoughtful limitations and ethics discussion. However, the headline comparisons are currently conditional on an unstated data-provenance assumption: depth and text are said to be generated by Depth Anything and LLaVA, but the manuscript never specifies which image those models are given. Since both are RGB-based models, the natural reading is that they are applied to the ground-truth visible face, which would give MTVDiff access to the target image at train and test time. This issue, together with several internal metric inconsistencies, prevents the results from being interpreted as stated.

major comments (3)
  1. [§4.1 and §5 (Limitations)] The paper says 'Both datasets are augmented with depth maps generated by Depth Anything [26] and textual descriptions from LLaVA [13]' but never states the input image to these models. Depth Anything is a monocular RGB depth estimator and LLaVA is an RGB vision-language model. If they are applied to the ground-truth visible face, then at train and test time MTVDiff is conditioned on information derived from the exact image it is asked to reconstruct, while thermal-only baselines (Axial-GAN, T2V-DDPM, BBDM, AT-DDPM) are not. This would invalidate the head-to-head comparisons in Tables 1 and 2 and would inflate the headline 'FID reductions of up to 48.3%' and 'Rank-1 improvements of up to 8.9%'. Please state explicitly whether depth/text are computed from the visible ground truth, from the thermal input, or from some other source. If they are target-derived, rerun all comparisons with auxi
  2. [Table 1 vs. Table 3 vs. Fig. 7] Three configurations that all appear to be thermal-only report different numbers on SpeakingFaces. Table 1 'MTVDiff (w/ Thermal)' reports FID 20.14, Table 3 Variant A reports FID 20.14, and Fig. 7 'MTVDiff*' (described in §4.7 as 'relying solely on thermal input') reports FID 14.69. The Table 1 footnote says 'w/ Thermal' uses thermal in both branches and no depth, but it does not state whether text is used; §4.7 says MTVDiff* removes both depth and text. If the difference is modality-dropout training, that must be stated explicitly. As written, the reader cannot tell which number corresponds to which configuration, and the claim that 'gains from architectural design rather than modality alone' is not supported by the reported numbers.
  3. [Table 4 (Prompt Sensitivity)] The prompt sensitivity analysis is not a controlled comparison. The 'No Text (baseline)' row reports FID 85.79, LPIPS 0.1918, PSNR 19.62, SSIM 0.7232 — identical to Variant A (no depth, no text) in Table 3 — and the 'Complete Description' row reports 86.13/0.1864/19.81/0.7335, identical to Variant C (text only, no depth) in Table 3. The full no-text system (Variant D, or Table 1 'w/o Text') has FID 75.38 on MCXFace. Thus the rows vary the presence of depth as well as the prompt, so the conclusions that 'irrelevant text actively harms generation' and that 'LLaVA descriptions match manual descriptions' are not supported. Please re-run this analysis with a fixed depth-conditioned model and vary only the prompt text.
minor comments (5)
  1. [§4.3, Table 1] The sentence 'MTVDiff (w/ Thermal), using only thermal input, still outperforms most baselines on SpeakingFaces' is hard to verify because the row label is ambiguous about whether text is present. Please clarify the exact modality configuration of each row.
  2. [§4.5, Table 4] The 'Irrelevant Text' row is not tied to any configuration in Table 3. Please report the full modality settings (e.g., depth on/off, cross-attention on/off) for every row in Table 4.
  3. [§3.2, Eq. (9)] The symbol x in Eq. (9) (C_{1x1}(x) + F_t·s_3 + F_d·b_3) is not defined clearly; it should be stated whether x is the earlier fused feature, a branch feature, or the original input.
  4. [References] There are minor reference/citation issues: ref. [24] appears to contain a typo in the author name ('Timotfe' instead of 'Timofte'), and some conference names are inconsistently capitalized. Please proofread.
  5. [§4.2] The implementation details state 500 epochs and 48 hours on 4×RTX 4090s, but no random seeds or number of runs are reported. Please state whether the reported metrics are single-run or averaged, and provide seeds for reproducibility.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: benchmark claims rest on external metrics and independent ablations; only minor self-citation to prior DiffTV/DiffX work.

full rationale

The paper's headline results are empirical measurements against fixed external datasets (MCXFace, SpeakingFaces) using public FID/LPIPS/PSNR/SSIM and ArcFace Rank-1 metrics, so the SOTA claim is not derived from a self-referential equation or from a fitted constant renamed as a prediction. The ablations (Tab. 3) evaluate the depth, text, and cross-attention components separately, including thermal-only and text-free variants, so the final result does not collapse to a single input by construction. The authors cite their own prior work (DiffTV [12], DiffX [23]) as baselines and as the origin of the gated-cross-attention design; these are architectural/baseline references, not load-bearing proof, and the paper supplies its own quantitative comparisons. The §4.1 sentence 'Both datasets are augmented with depth maps generated by Depth Anything [26] and textual descriptions from LLaVA [13]' is ambiguous about the conditioning-image source, which could undermine the fairness of the comparison if the visible ground truth were used, but no statement in the paper exhibits the specific reduction required to call this circular (e.g., no equation showing the target is an input). That concern belongs to correctness/validity, not circularity. Overall: no circular derivation chain; minor self-citation only.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard diffusion training, on the availability and accuracy of estimated depth and text conditions, and on several hand-chosen hyperparameters. No new physical entities are introduced; DBCAF and gated alignment are architectural modules, not entities with independent falsifiable predictions.

free parameters (3)
  • λ gating scale = not reported
    Appears in Eqs. (12)–(15) as a global multiplier for gated cross-attention, FFN, self-attention, and MLP modulation. If set by hand, it directly controls how strongly text/depth influence generation; the value is not given.
  • modality dropout probabilities = p_depth=0.1, p_thermal=0.1, p_text=0.5
    Chosen in §4.7 to reflect assumed acquisition reliability. These probabilities shape the robustness claim and are not justified by sensitivity analysis.
  • SFT injection resolutions = 64x64, 32x32, 16x16
    In §3.4, three SFT injection points are selected to match DBCAF scales; alternative injection placements or scales are not ablated.
assumptions (4)
  • standard math The DDPM/LDM denoising objective (Eq. 2) is a valid training objective for conditional image synthesis.
    Standard from Ho et al. and Rombach et al.; not proved here, but standard background.
  • domain assumption Depth maps from Depth Anything accurately represent facial geometry and are available at inference from an RGB-D sensor.
    §4.1: depth maps are generated rather than captured, and no depth-quality validation is provided. If depth is estimated from the visible target image, this assumption is violated.
  • domain assumption LLaVA-generated structured text captures the identity-relevant attributes needed for semantic guidance, and CLIP text features align with visual features.
    §3.3 and §4.5: text is generated from an unspecified input; the prompt sensitivity table shows wrong text hurts performance, but LLaVA error rates are not analyzed.
  • domain assumption A frozen Stable Diffusion backbone provides a suitable generative prior for face synthesis at 256×256 with only encoder/SFT adaptation.
    §3.4: SD weights are kept frozen; the paper does not compare against fine-tuning the backbone or training from scratch.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MTVDiff: Multimodal Conditional Latent Diffusion for Enhanced Thermal-to-Visible Face Translation." pith.science (2026). https://pith.science/paper/XLVH72UP

@misc{pith2026260719886,
  author       = {Pith},
  title        = {Pith review of: MTVDiff: Multimodal Conditional Latent Diffusion for Enhanced Thermal-to-Visible Face Translation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XLVH72UP}},
  note         = {Machine review of arXiv:2607.19886}
}
read the original abstract

Thermal-to-visible face translation presents fundamental challenges including geometric discontinuities, semantic attribute mismatches, and identity degradation. We propose MTVDiff, a novel multimodal latent diffusion framework that synergistically integrates depth and textual information to address these limitations while preserving identity characteristics. The MTVDiff framework presents three core technical contributions: (1) a Dual-Branch Cross-Attention Fusion (DBCAF) module for multi-scale thermal-depth feature extraction and fusion; (2) a Gated Text-to-Visual Feature Alignment mechanism for semantically-guided generation; and (3) Spatial Feature Transformations (SFT) for adaptive multimodal prior integration. Extensive experiments on the MCXFace and SpeakingFaces datasets demonstrate that our multimodal approach significantly outperforms existing GAN-based and diffusion-based approaches across multiple metrics, achieving substantial improvements in both image quality and face verification performance, with FID reductions of up to 48.3% and Rank-1 accuracy improvements of up to 8.9\%. Our work provides a robust solution for face recognition systems operating under varying illumination conditions and advances the state-of-the-art in cross-spectral facial image translation through effective multimodal integration.

Figures

Figures reproduced from arXiv: 2607.19886 by the authors.

Figure 1
Figure 1. The overall framework of the proposed multi-modal conditional latent diffusion (MTVDiff). 2 Related Work 2.1 Image-to-Image Translation GAN-based Approaches. Early image-to-image (I2I) translation methods predominantly employed Generative Adversarial Networks (GANs) [9, 30]. Spe￾cialized architectures like Axial-GAN [8] or pix2pix-zero [17] integrated attention mechanisms or distillation for cross-modal translation … view at source ↗
Figure 2
Figure 2. Illustration of the Dual-Branch Cross-Attention Fusion (DBCAF) module. 3.3 Gated Text-to-Visual Feature Alignment To leverage semantic information from textual descriptions, we encode textual descriptions into embedding tensors via CLIP’s text encoder. Although standard cross-attention mechanisms in Stable Diffusion effectively incorporate text condi￾tions, we identify limitations in the interaction between visual l… view at source ↗
Figure 3
Figure 3. Comparative visualization on the SpeakingFaces dataset across gender and eth￾nicity. (a) Thermal, (b) Axial-GAN, (c) Uni-ControlNet, (d) BBDM, (e) T2V-DDPM, (f) DiffV2IR, (g) DiffTV, (h) Ours, (i) GT. Face Recognition Comparison [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: and [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Ablation study visualization on the SpeakingFaces dataset. (a) Thermal, (b) Baseline, (c) Text, (d) Depth, (e) D+T, (f) D+CA, (g) MTVDiff, (h) GT. 4.5 Text Prompt Sensitivity Analysis [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Temporal evolution of gating parameters across diffusion timesteps. 4.6 Temporal Evolution of Gating Parameters We analyze how the six gating parameters (s1, s2, s3, b1, b2, b3) evolve across dif￾ferent diffusion timesteps [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Performance comparison between DiffTV, MTVDiff* (thermal-only), and MTVDiff (full multimodal) on SpeakingFaces. 5 Conclusion We presented MTVDiff, a multimodal latent diffusion framework for thermal￾to-visible face translation that synergistically integrates thermal im…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

30 extracted references · 3 linked inside Pith

  1. [26]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Yang, L., Kang, B., Huang, Z., Xu, X., Feng, J., Zhao, H.: Depth anything: Un- leashing the power of large-scale unlabeled data. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 10371–10381 (2024)

  2. [13]

    Advances in neural information processing systems36, 34892–34916 (2023)

    Liu, H., Li, C., Wu, Q., Lee, Y.J.: Visual instruction tuning. Advances in neural information processing systems36, 34892–34916 (2023)

  3. [1]

    Sensors21(10), 3465 (2021)

    Abdrakhmanova, M., Kuzdeuov, A., Jarju, S., Khassanov, Y., Lewis, M., Varol, H.A.: Speakingfaces: A large-scale multimodal dataset of voice commands with visual and thermal video streams. Sensors21(10), 3465 (2021)

  4. [2]

    In: 16 Z

    Anghelone, D., Chen, C., Faure, P., Ross, A., Dantcheva, A.: Explainable thermal to visible face recognition using latent-guided generative adversarial network. In: 16 Z. Xia et al. 2021 16th IEEE International Conference on Automatic Face and Gesture Recog- nition (FG 2021). pp. 1–8. IEEE (2021)

  5. [3]

    Neurocomputing611, 128626 (2025)

    Anghelone, D., Chen, C., Ross, A., Dantcheva, A.: Beyond the visible: A survey on cross-spectral face recognition. Neurocomputing611, 128626 (2025)

  6. [4]

    arXiv preprint arXiv:1701.04862 (2017)

    Arjovsky, M., Bottou, L.: Towards principled methods for training generative ad- versarial networks. arXiv preprint arXiv:1701.04862 (2017)

  7. [5]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Deng, J., Guo, J., Xue, N., Zafeiriou, S.: Arcface: Additive angular margin loss for deep face recognition. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 4690–4699 (2019)

  8. [6]

    IEEE Transactions on Infor- mation Forensics and Security18, 133–146 (2022)

    George, A., Mohammadi, A., Marcel, S.: Prepended domain transformer: Hetero- geneous face recognition without bells and whistles. IEEE Transactions on Infor- mation Forensics and Security18, 133–146 (2022)

Show all 30 references
  1. [7]

    Advances in neural information processing systems33, 6840–6851 (2020)

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in neural information processing systems33, 6840–6851 (2020)

  2. [8]

    In: 2021 IEEE Interna- tional Joint Conference on Biometrics (IJCB)

    Immidisetti, R., Hu, S., Patel, V.M.: Simultaneous face hallucination and transla- tion for thermal to visible face verification using axial-gan. In: 2021 IEEE Interna- tional Joint Conference on Biometrics (IJCB). pp. 1–8. IEEE (2021)

  3. [9]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Isola, P., Zhu, J.Y., Zhou, T., Efros, A.A.: Image-to-image translation with condi- tional adversarial networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 1125–1134 (2017)

  4. [10]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern Recognition

    Li, B., Xue, K., Liu, B., Lai, Y.K.: Bbdm: Image-to-image translation with brow- nian bridge diffusion models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern Recognition. pp. 1952–1961 (2023)

  5. [11]

    IEEE Signal Processing Letters (2024)

    Lin, J., Wu, Y., Wang, Z., Liu, X., Guo, Y.: Pair-id: A dual modal framework for identity preserving image generation. IEEE Signal Processing Letters (2024)

  6. [12]

    In: Proceedings of the 32nd ACM International Confer- ence on Multimedia

    Lin, J., Zhao, G., Xu, J., Wang, G., Wang, Z., Dantcheva, A., Du, L., Chen, C.: Difftv: Identity-preserved thermal-to-visible face translation via feature alignment and dual-stage conditions. In: Proceedings of the 32nd ACM International Confer- ence on Multimedia. pp. 10930–1...

  7. [14]

    In: Proceedings of the 32nd ACM International Conference on Multimedia

    Lu, X., Hu, X., Luo, J., Zhu, B., Ruan, Y., Ren, W.: 3d priors-guided diffusion for blind face restoration. In: Proceedings of the 32nd ACM International Conference on Multimedia. pp. 1829–1838 (2024)

  8. [15]

    In: Proceedings of the AAAI conference on artificial intelligence

    Mou, C., Wang, X., Xie, L., Wu, Y., Zhang, J., Qi, Z., Shan, Y.: T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. In: Proceedings of the AAAI conference on artificial intelligence. vol. 38, pp. 4296–4304 (2024)

  9. [16]

    In: 2023 IEEE 17th International Conference on Automatic Face and Gesture Recognition (FG)

    Nair, N.G., Patel, V.M.: T2v-ddpm: Thermal to visible face translation using de- noising diffusion probabilistic models. In: 2023 IEEE 17th International Conference on Automatic Face and Gesture Recognition (FG). pp. 1–7. IEEE (2023)

  10. [17]

    In: ACM SIGGRAPH 2023 conference proceedings

    Parmar, G., Kumar Singh, K., Zhang, R., Li, Y., Lu, J., Zhu, J.Y.: Zero-shot image- to-image translation. In: ACM SIGGRAPH 2023 conference proceedings. pp. 1–11 (2023)

  11. [18]

    arXiv preprint arXiv:2503.19012 (2025)

    Ran, L., Wang, L., Wang, G., Wang, P., Zhang, Y.: Diffv2ir: visible-to-infrared dif- fusion model via vision-language understanding. arXiv preprint arXiv:2503.19012 (2025)

  12. [19]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 10684–10695 (2022) MTVDiff: Multimodal Conditional Latent Diffusion 17

  13. [20]

    arXiv preprint arXiv:2510.22994 (2025)

    Song, Q., Zhou, D., Lin, J., Shen, F., Wang, J., Hu, X., Chen, C., Heng, P.A.: Scenedecorator: Towards scene-oriented story generation with scene planning and scene consistency. arXiv preprint arXiv:2510.22994 (2025)

  14. [21]

    Advances in neural information processing systems30(2017)

    Van Den Oord, A., Vinyals, O., et al.: Neural discrete representation learning. Advances in neural information processing systems30(2017)

  15. [22]

    International Journal of Computer Vision 132(12), 5929–5949 (2024)

    Wang, J., Yue, Z., Zhou, S., Chan, K.C., Loy, C.C.: Exploiting diffusion prior for real-world image super-resolution. International Journal of Computer Vision 132(12), 5929–5949 (2024)

  16. [23]

    arXiv preprint arXiv:2407.15488 (2024)

    Wang, Z., Lin, J., Qian, Y., Huang, Y., Tian, S., Chai, B., Deng, J., Yang, Q., Du, L., Chen, C., et al.: Diffx: Guide your layout to cross-modal generative modeling. arXiv preprint arXiv:2407.15488 (2024)

  17. [24]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

    Xia, B., Zhang, Y., Wang, S., Wang, Y., Wu, X., Tian, Y., Yang, W., Timotfe, R., Van Gool, L.: Diffi2i: Efficient diffusion model for image-to-image translation. IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

  18. [25]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

    Xia, M., Zhou, Y., Yi, R., Liu, Y.J., Wang, W.: A diffusion model translator for efficient image-to-image translation. IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

  19. [27]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Zhang, L., Rao, A., Agrawala, M.: Adding conditional control to text-to-image diffusion models. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 3836–3847 (2023)

  20. [28]

    Zhao, S., Chen, D., Chen, Y.C., Bao, J., Hao, S., Yuan, L., Wong, K.Y.K.: Uni- controlnet:All-in-onecontroltotext-to-imagediffusionmodels.AdvancesinNeural Information Processing Systems (2023)

  21. [29]

    arXiv preprint arXiv:2512.23519 (2025)

    Zhou, D., Lin, J., Shen, G., Liu, Q., Gao, J., Liu, L., Du, L., Chen, C., Fu, C.W., Hu, X., et al.: Identitystory: Taming your identity-preserving generator for human- centric story generation. arXiv preprint arXiv:2512.23519 (2025)

  22. [30]

    In: Proceedings of the IEEE interna- tional conference on computer vision

    Zhu, J.Y., Park, T., Isola, P., Efros, A.A.: Unpaired image-to-image translation using cycle-consistent adversarial networks. In: Proceedings of the IEEE interna- tional conference on computer vision. pp. 2223–2232 (2017)

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.